Installation

• Currently, kszx works on Linux but not Apple! If you need Apple support, let me know and I can prioritize.

• Step 1. (Optional but recommended.) Install some prerequisites with conda.

  This may not be necessary, but maximizes the probability that the pip install in step 2 will work. (If you skip step 1, then pip install will install a long chain of dependencies, some of which have hidden dependence on tools installed on your system, e.g. compilers, automake, libtool.)

  You can either install the “precomputed” conda env in kszx/environment.yml:

  git clone https://github.com/kmsmith137/kszx
  cd kszx
  conda env create -n kszx -f environment.yml
  conda activate kszx
  

  Or build your own conda env with:

   # Note: gcc 14 conflicts with camb (or at least conda thinks so), so force gcc==13.
   # Note: jupyterlab is only needed if you're running jupyterlab locally.
   # Note: sphinx packages are only needed if you're building sphinx docs locally.
   # Note: emcee/corner are only needed for MCMCs with 'class kszx.PgvLikelihood'.
  
   conda create -n kszx -c conda-forge \
     gxx_linux-64==13.3.0 \
     python==3.12.6 pybind11 \
     jupyterlab \
     sphinx sphinx-math-dollar \
     h5py python-wget astropy camb fitsio healpy pixell \
     emcee corner
  
  conda activate kszx
  

  The second approach takes longer, but may be preferred if you want to customize/extend the environment.

  (Note: if you don’t have conda installed, I recommend the miniforge version of conda, available at https://github.com/conda-forge/miniforge, table heading “Miniforge3”.)

• Step 2: Install kszx with pip. Do one of the following:

  • My preferred ‘pip install’ is an “editable install” (pip install -e), which is nice because you can edit source files without doing ‘pip install’ again:

    # Editable install
    git clone https://github.com/kmsmith137/kszx
    cd kszx
    pip install -v -e .
    

  • For a non-editable install, omit the -e flag:

    # Non-editable install
    git clone https://github.com/kmsmith137/kszx
    cd kszx
    pip install -v .
    

  • Finally, there is a kszx version on PyPI (https://pypi.org/project/kszx/), so you can install without cloning the github repo, but it may not be the most current:

    # PyPI install (may not be current)
    pip install -v kszx
    

• Step 3: I recommend testing that the install worked, with:

  # Just runs some unit tests
  python -m kszx test
  

• Step 4 (optional): install Mat Madhavacheril’s hmvec library. This library isn’t currently used in kszx itself, but is used in some of the notebooks (https://github.com/kmsmith137/kszx_notebooks/):

  # Optional: install hmvec
  pip install -v git+https://github.com/simonsobs/hmvec
  

• Step 5: You’ll probably want to use kszx from a jupyter notebook, so you’ll want to make sure you can import kszx from jupyter.

  The details of the setup will depend on how you’re running jupyter, so it’s hard for me to give general instructions, but if you need help let me know. If you’re running jupyterlab from the command line yourself, then a straightforward approach is to simply run the jupyterlab executable from an environment that has kszx installed. (Note that the conda env in step 1 contains jupyterlab already, but if you skipped step 1, then you’ll need to do pip install jupyterlab.)